Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

This paper argues that a shallow quantum neural network, trained on real daily weather data, can match or beat a classical recurrent baseline in short- and medium-term forecasting, with the strongest edge in wind speed.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-05 12:33 UTC pith:7IU6BD3W

load-bearing objection A small but honest QML weather-forecasting study whose central comparison is plausible but not yet established; the missing feature-date alignment is the thing that needs fixing. the 4 major comments →

arxiv 2509.01422 v1 pith:7IU6BD3W submitted 2025-09-01 quant-ph cs.LG

Exploring Quantum Machine Learning for Weather Forecasting

Maria Helo\'isa F. da Silva , Gleydson F. de Jesus , Christiano M. S. Nascimento , Val\'eria L. da Silva , Clebson Cruz This is my paper
classification quant-ph cs.LG
keywords quantum machine learningweather forecastingquantum neural networkvariational quantum circuitregressionwind speed predictiontemperature predictionPOWER dataset
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper asks whether a quantum neural network can do real forecasting work, not just classification. Using a year of daily weather data from the POWER database for a Brazilian city, the authors train six small variational quantum circuits and one classical recurrent network to predict temperature 14 days ahead and wind speed 5 days ahead. Their best quantum configurations achieve mean absolute errors of 0.304 (temperature) and 0.156 (wind speed), versus 0.357 and 0.167 for the classical network. The authors read this as evidence that QNNs can be competitive in short- and medium-term climate prediction and may adapt to abrupt shifts better than a classical RNN.

Core claim

On the paper's own terms, the central discovery is that a low-depth variational quantum circuit can be trained on real meteorological data to forecast temperature and wind speed with accuracy on par with—and in the best configurations better than—a classical recurrent baseline. The quantum model's forecast distributions sit closer to observed values on days with sharp weather changes, which the authors attribute to the QNN's ability to capture nonlinear patterns without long-range memory. They also find that network depth and entanglement strategy affect performance nonlinearly, so the best configuration differs between temperature and wind speed.

What carries the argument

The load-bearing object is a parameterized quantum circuit: each selected climate feature is encoded as a Ry rotation on one qubit, the qubits are entangled with either a simple entangling layer or a strongly entangling layer, and trainable rotation angles are updated by a classical optimizer after projective measurement. The paper varies only the number of variational layers (1, 3, or 5) and the entanglement strategy, holding data, optimizer, and training budget fixed, which isolates the architectural effect on forecasting skill.

Load-bearing premise

The load-bearing premise is that every input feature is known before the forecast horizon starts; if humidity, solar irradiance, or a lagged variable is measured on the target day, the reported mean absolute errors are not genuine forecasts.

What would settle it

Look at the timestamps of the six (temperature) or eight (wind) features used for the test days in the POWER data. If any feature value falls on or after the first forecast day, re-run the comparison with features shifted so all inputs predate the horizon; the claimed QNN edge would disappear if those same-day features carried the signal.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If the reported results hold, quantum neural networks deserve inclusion in short- and medium-term regression benchmarks, not just classification tasks.
  • The QNN's better handling of abrupt temperature and wind shifts suggests quantum models could be tested for extreme-event forecasting, the regime where classical models are weakest.
  • Because shallow circuits matched or beat deeper ones, future quantum forecasting work should focus on circuit structure rather than brute-force depth.
  • The faster convergence observed for temperature suggests QNNs may be a practical choice when training budgets are tight.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The comparison is only meaningful if every input feature is known before the forecast window begins; the paper never states this temporal alignment, so the reported edge over the RNN is conditional on that unstated assumption.
  • A natural extension is a strict forecast test where features are shifted backward by the full horizon, so no same-day information leaks into the predictors; this would tell whether the QNN advantage survives a true out-of-sample forecast.
  • The same architecture could be stress-tested on multi-year data from other climates, where abrupt shifts are more frequent, to see whether the adaptability claim generalizes.
  • Because the test sets are tiny (14 and 5 days), the MAE differences lie within plausible sampling noise; a larger test window is needed to confirm the quantum edge.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper reports a proof-of-concept comparison between a variational Quantum Neural Network (QNN) and a classical Recurrent Neural Network (RNN) for short/medium-term weather forecasting at a single Brazilian location, using one year of daily NASA POWER data. Six QNN configurations (two entanglement strategies, three depths) are compared against an RNN baseline for 14-day temperature and 5-day wind-speed forecasting. The reported test-set MAEs favor the best QNN configuration (0.304 vs 0.357 for temperature, 0.156 vs 0.167 for wind speed), and the authors further claim qualitative advantages in adaptability to abrupt shifts and in training speed for temperature. The manuscript includes reproducibility information and cites a public code repository.

Significance. If the central claim were established, the paper would provide evidence that small variational QNNs can be competitive with classical recurrent models on structured meteorological time series, and it would add a useful data point to the sparse literature on QML for regression. The authors deserve credit for using real meteorological data, for running 10 repetitions per configuration, for reporting training/validation loss curves, and for making source code available in a public repository (reference [42]). However, the quantitative evidence is currently not strong enough: the test sets contain only 14 and 5 points, the best configuration is selected on the same test set with no multiplicity correction or uncertainty interval, the classical baseline is trained with different hyperparameters and much longer epoch counts, and the temporal alignment of input features to forecast targets is not specified. These issues are load-bearing because the gap to the RNN is small and could be explained by leakage, tuning artifacts, or noise. With the requested clarifications and a more careful statistical analysis, the result would be a solid, if modest, contribution; in its present form it is an

major comments (4)
  1. [II B, III A, III B, Tables III and V] The paper never states the temporal alignment of the input features to the forecast targets. Section II B only says that lags of 28 days (temperature) and 6 days (wind speed) were appended; it does not say whether the five or seven climate variables (humidity, solar irradiance, etc.) are values from the forecast origin, from the target date, or from some other date. Tables III and V list only the number of features, not their dates relative to the horizon. Since the test horizon is 14 or 5 days, any feature observed inside that window would make the reported MAEs nowcasts rather than forecasts and would invalidate the comparison. Please specify precisely, for each sample, which calendar dates are used for each input feature and which date is the forecast target, and confirm that all inputs are known before the forecast window begins. The code in repository [42] should make this explicit.
  2. [II C, III A 2, III B 2] The central quantitative claim rests on test MAEs computed on 14 and 5 test points. No uncertainty intervals are reported for the MAE values; Table IV and Table VI give only point estimates for each configuration, despite the paper stating that 10 runs were performed. With test sizes this small, the differences (0.304 vs 0.357 and 0.156 vs 0.167) may be within run-to-run and sampling variability. In addition, the best QNN configuration is selected after inspecting the same test set on which it is then reported, which introduces selection bias and inflates the apparent advantage. Please report the distribution of the 10 per-run MAEs, provide standard errors or confidence intervals, and, if the best configuration is chosen from the six, account for the multiple comparisons (e.g., by reporting all configurations and applying a simple correction, or by using a validation split for model sele
  3. [II A, III A, III B] The comparison between QNN and RNN is not controlled: the QNN uses learning rate 0.1 and 30 epochs, while the RNN uses learning rate 0.001 and 500 epochs (Tables III and V and the text in Sections III A 2 and III B 2). The RNN also uses 256 neurons, whereas the QNN uses 6 or 8 qubits. Differences in optimization budget and model capacity are confounded with the quantum/classical distinction. At minimum, the authors should either tune the RNN to a similar practical training budget, provide a sensitivity analysis of the RNN to learning rate and epoch count, or explicitly state that the comparison is between one particular QNN implementation and one particular RNN implementation, and temper the abstract's language accordingly.
  4. [II B, III A, III B] The evaluation uses a single train/test split at the end of a one-year series, with no seasonal coverage (10 days for temperature, 5 days for wind). This design is particularly fragile for weather data, which is strongly seasonal and autocorrelated. The claim of 'abrupt data shift' adaptability is based on a single day (day 11 in temperature, day 1 in wind) in each split. Please add rolling-origin evaluation or at least multiple test windows, and avoid drawing general conclusions about 'abrupt shifts' from one or two points.
minor comments (5)
  1. [Section III A 2] The text says 'the MAE of 0.347 achieved by the RNN' immediately after presenting Table IV, which reports 0.357. Please correct the inconsistency.
  2. [Figures 9-12, 20] Several figure cross-references in Section III are mislabeled: the classical temperature mean prediction is referred to as 'Figure 19' instead of Figure 11, and the classical wind loss is referred to as 'Figure 20' instead of Figure 12/20 depending on section. Figure 14's caption says 'daily temperature data' but the panel shows wind speed. Please renumber and cross-check all figure references.
  3. [Figure 5] The caption refers to 'the presented INN' where it presumably means 'QNN' or 'VQA'. Please fix the typo.
  4. [Eq. (1), Tables IV and VI] The data are standardized with Eq. (1), so the reported MAE values are in standardized units, not in °C or m/s. The manuscript does not state this, which makes the numbers hard to interpret. Please clarify in the metric definitions or in the table captions.
  5. [Reference [42]] The repository reference is incomplete: it says 'gitHub repository' without a URL or identifier. A complete link is needed to support the reproducibility claim.

Circularity Check

0 steps flagged

No circular reduction; only minor self-citation for QNN architecture, not load-bearing.

full rationale

The paper's claimed derivation is an empirical benchmark: a QNN architecture (encoding six/eight features, parameterized rotations, measured and classically optimized) is trained on NASA POWER data and compared to an RNN under matched train/test splits; the reported MAEs (0.304 vs 0.357 for temperature, 0.156 vs 0.167 for wind) are computed on held-out terminal segments and are not algebraic functions of the training inputs or of the fitted parameters. The only self-citation is the architectural lineage: 'designed following the approach proposed by Oğur [38] and further explored by Jesus [39]' (Sec. II C 1), where [39] shares authors with this paper. That citation supplies the ansatz (Ry encoding + variational layers + entanglement strategy), but it does not by itself determine the comparative result; the same architecture could have lost, and the empirical outcomes are externally grounded in the dataset. I therefore identify no circular step: no quantity called a prediction is identical by construction to a fitted input, no uniqueness theorem is imported from the authors' prior work, and no known result is merely renamed. The unspecified temporal alignment of features is a validity risk (if covariates come from inside the forecast window, the experiment is nowcasting), but that would be data leakage, not circularity; it does not affect this score.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The central claim is an empirical comparison, so the ledger lists the design choices and assumptions the comparison rests on. No new physical entities are introduced.

free parameters (5)
  • Lag periods = 28 days (temperature), 6 days (wind)
    Chosen from Pearson correlations on the same dataset (Section II B), not independently validated.
  • Feature correlation threshold = |rho| >= 0.3
    Hand-chosen threshold for retaining features (Section II B).
  • QNN hyperparameters = lr=0.1, 30 epochs, batch size 10
    These differ from the RNN baseline (lr=0.001, 500 epochs), confounding the comparison (Section III A/B).
  • Best architecture selection = Exp 2, 1 layer (temp); Exp 1, 3 layers (wind)
    Selected via minimum test-set MAE after the fact (Tables IV and VI), inducing selection bias.
  • Validation split and seeds = 0.1 split; seeds not reported
    Not documented, complicating reproducibility.
axioms (4)
  • domain assumption NASA POWER daily weather data for Barreiras is an adequate ground-truth source
    Adopted in Section II A; data quality is assessed qualitatively in Table I.
  • domain assumption Pearson correlation is a valid guide to predictive features and lags for a nonlinear model
    Used for feature and lag selection in Section II B; a linear screening assumption.
  • ad hoc to paper The variational QNN architecture of Oğur [38] and Jesus [39] is appropriate for time-series regression
    Used in Section II C 1 without theoretical justification.
  • ad hoc to paper Input features are observed before the forecast horizon (no data leakage)
    Unstated in Section II B/C; critical for the validity of the comparison.

pith-pipeline@v1.4.0-alltime-deepseek-medium · 10970 in / 16862 out tokens · 172311 ms · 2026-08-05T12:33:16.299972+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Exploring Quantum Machine Learning for Weather Forecasting." pith.science (2026). https://pith.science/paper/7IU6BD3W

@misc{pith2026250901422,
  author       = {Pith},
  title        = {Pith review of: Exploring Quantum Machine Learning for Weather Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7IU6BD3W}},
  note         = {Machine review of arXiv:2509.01422}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Weather forecasting plays a crucial role in supporting strategic decisions across various sectors, including agriculture, renewable energy production, and disaster management. However, the inherently dynamic and chaotic behavior of the atmosphere presents significant challenges to conventional predictive models. On the other hand, introducing quantum computing simulation techniques to the forecasting problems constitutes a promising alternative to overcome these challenges. In this context, this work explores the emerging intersection between quantum machine learning (QML) and climate forecasting. We present the implementation of a Quantum Neural Network (QNN) trained on real meteorological data from NASA's Prediction of Worldwide Energy Resources (POWER) database. The results show that QNN has the potential to outperform a classical Recurrent Neural Network (RNN) in terms of accuracy and adaptability to abrupt data shifts, particularly in wind speed prediction. Despite observed nonlinearities and architectural sensitivities, the QNN demonstrated robustness in handling temporal variability and faster convergence in temperature prediction. These findings highlight the potential of quantum models in short and medium term climate prediction, while also revealing key challenges and future directions for optimization and broader applicability.

Figures

Figures reproduced from arXiv: 2509.01422 by Christiano M. S. Nascimento, Clebson Cruz, Gleydson F. de Jesus, Maria Helo\'isa F. da Silva, Val\'eria L. da Silva.

Figure 1
Figure 1. Figure 1: FIG. 1: Pearson correlation matrix for the selected meteorological parameters. The interpretation of the Pearson [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2: Pearson correlation between temperature and [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3: Pearson correlation between wind speed and [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4: Operating principle of the variational quantum algorithm (VQA). Classical inputs—here, temperature and [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5: Entanglement strategies applied to the pre [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7: Subdivision of the dataset into training set [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8: Distribution of quantum temperature forecasts for 14 days ahead, averaged over 10 runs per experiment. [PITH_FULL_IMAGE:figures/full_fig_p007_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9: Average training loss (blue) and validation loss (red) values for quantum temperature forecasting 14 days [PITH_FULL_IMAGE:figures/full_fig_p008_9.png] view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12: Average loss function values (blue) and valida [PITH_FULL_IMAGE:figures/full_fig_p009_12.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11: Distribution of classical temperature forecast [PITH_FULL_IMAGE:figures/full_fig_p009_11.png] view at source ↗
Figure 16
Figure 16. Figure 16 [PITH_FULL_IMAGE:figures/full_fig_p009_16.png] view at source ↗
Figure 14
Figure 14. Figure 14: FIG. 14: Actual daily wind speed data obtained from [PITH_FULL_IMAGE:figures/full_fig_p010_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: FIG. 15: Subdivision of the daily wind speed dataset [PITH_FULL_IMAGE:figures/full_fig_p010_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: FIG. 16: Distribution of 5-day-ahead wind forecast data based on the average of 10 runs per experiment. Subfigures [PITH_FULL_IMAGE:figures/full_fig_p011_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: FIG. 17: Average values of the training loss (blue) and validation loss (red) over 30 training epochs, averaged across [PITH_FULL_IMAGE:figures/full_fig_p012_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: FIG. 18: Distribution of classical wind speed forecasts [PITH_FULL_IMAGE:figures/full_fig_p013_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: FIG. 19: Mean classical wind speed forecast for 5 days [PITH_FULL_IMAGE:figures/full_fig_p013_19.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. How Hard Is Quantum Advantage? A Cloud Microphysics Stress Test for Variational Quantum Models

    quant-ph 2026-07 conditional novelty 5.0

    Even after rich frequency encoding, classical post-processing, and extensive hyperparameter search, hybrid QNNs underperform simple FCNNs on cloud microphysics parameterization.

Reference graph

Works this paper leans on

48 extracted references · 46 canonical work pages · cited by 1 Pith paper

  1. [1]

    Quantum Neural Network (QNN) The quantum predictive model employed in this study is a Quantum Neural Network (QNN), designed following the approach proposed by O˘ gur [38] and further explored by Jesus [39] in demand forecasting for the financial sec- tor. The operating principle of the underlying variational quantum algorithm—where classical features xi ...

    work page

  2. [2]

    While this ar- chitecture has demonstrated effectiveness in time-series forecasting tasks, its ability to capture abrupt temporal variations is comparatively limited [41]

    Recurrent Neural Network (RNN) The classical predictive model adopted in this study was a Recurrent Neural Network (RNN), implemented following the approach described by [40]. While this ar- chitecture has demonstrated effectiveness in time-series forecasting tasks, its ability to capture abrupt temporal variations is comparatively limited [41]. III. RESU...

    work page 2023

  3. [3]

    Since we selected five climate vari- ables along with the time-lagged target variable, the net- work architecture uses a total of six qubits, each encoding one feature

    Quantum Model The second hyperparameter defined was the number of qubits in the QNN. Since we selected five climate vari- ables along with the time-lagged target variable, the net- work architecture uses a total of six qubits, each encoding one feature. Table III presents the experimental setup used for the temperature forecasting tasks. Results were obta...

    work page

  4. [4]

    However, due to its classical nature, incompatible with the use of qubits, 256 neurons were employed

    Classical Model The RNN used for classical temperature forecast- ing was configured with the same number of features, validation split, batch size, hardware, and training and testing proportions as the QNN, as shown in Table III. However, due to its classical nature, incompatible with the use of qubits, 256 neurons were employed. Another modified hyperpar...

    work page 2023

  5. [5]

    Validation loss Epochs Mean Training and Validation Loss Training loss (b) Experiment 2 – 1 layer

    Quantum Model The second hyperparameter defined was the number of qubits in the QNN, as seven climate variables were previ- ously selected, along with the time-lagged target variable; 8 Validation loss Epochs Mean Training and Validation Loss Training loss (a) Experiment 1 – 1 layer. Validation loss Epochs Mean Training and Validation Loss Training loss (...

    work page

  6. [6]

    iNOV ATeQ Lato Sensu Specialization in Quan- tum Computing – Researcher

    Classical Model The RNN used for classical wind speed forecast- ing was configured with the same number of fea- Validation loss Epochs Mean Training and Validation Loss Training loss FIG. 12: Average loss function values (blue) and valida- tion loss values (red) for classical temperature forecasting 14 days ahead, calculated as the average over 10 runs. A...

    work page 2023

  7. [7]

    B.-W. Shen, R. Pielke Sr, X. Zeng, J. Cui, S. Faghih- Naini, W. Paxson, A. Kesarkar, X. Zeng, and R. Atlas, Atmosphere 13, 1892 (2022)

    work page 2022

  8. [8]

    R. M. Pujahari, S. P. Yadav, and R. Khan, inApplication of Machine Learning in Agriculture (Elsevier, 2022) pp. 113–130

    work page 2022

  9. [9]

    Schultz, S

    M. Schultz, S. Reitmann, and S. Alam, Transportation Research Part C: Emerging Technologies 131, 103119 (2021)

    work page 2021

  10. [10]

    Wang and Y

    Y. Wang and Y. Zhang, Transportation Research Part C: Emerging Technologies 125, 103049 (2021)

    work page 2021

  11. [11]

    J. Li, L. Xu, L. Tang, S. Wang, and L. Li, Tourism management 68, 301 (2018)

    work page 2018

  12. [12]

    Enr ´ ıquez, M

    R. Enr ´ ıquez, M. J. Jim´ enez, and M. del Rosario Heras, Energy Procedia 91, 1024 (2016)

    work page 2016

  13. [13]

    L. M. Braman, M. K. van Aalst, S. J. Mason, P. Suarez, Y. Ait-Chellouche, and A. Tall, Disasters 37, 144 (2013)

    work page 2013

  14. [14]

    G. J. Van Oldenborgh, H. Hendon, T. Stockdale, M. L’Heureux, E. C. De Perez, R. Singh, and M. Van Aalst, Environmental research letters 16, 044003 (2021)

    work page 2021

  15. [15]

    Bochenek and Z

    B. Bochenek and Z. Ustrnul, Atmosphere 13, 180 (2022)

    work page 2022

  16. [16]

    Meenal, D

    R. Meenal, D. Binu, K. Ramya, P. A. Michael, K. Vinoth Kumar, E. Rajasekaran, and B. Sangeetha, Archives of Computational Methods in Engineering 29, 2875 (2022)

    work page 2022

  17. [17]

    Spiridonov, M

    V. Spiridonov, M. ´Curi´ c, V. Spiridonov, and M. ´Curi´ c, Fundamentals of Meteorology , 349 (2021)

    work page 2021

  18. [18]

    J. A. Brotzge, D. Berchoff, D. L. Carlis, F. H. Carr, R. H. Carr, J. J. Gerth, B. D. Gross, T. M. Hamill, S. E. Haupt, N. Jacobs, et al., Bulletin of the American Meteorological Society 104, E698 (2023)

    work page 2023

  19. [19]

    Zeguendry, Z

    A. Zeguendry, Z. Jarir, and M. Quafafou, Entropy 25, 287 (2023)

    work page 2023

  20. [20]

    S. S. Gill, A. Kumar, H. Singh, M. Singh, K. Kaur, M. Us- man, and R. Buyya, Software: Practice and Experience 52, 66 (2022)

    work page 2022

  21. [21]

    I. D. Lins, L. M. M. Araujo, C. B. S. Maior, P. M. da Silva Ramos, M. J. das Chagas Moura, A. J. Ferreira- Martins, R. Chaves, and A. Canabarro, Process Safety and Environmental Protection 186, 1197 (2024)

    work page 2024

  22. [22]

    Cerezo, G

    M. Cerezo, G. Verdon, H.-Y. Huang, L. Cincio, and P. J. Coles, Nature Computational Science 2, 567 (2022)

    2022

  23. [23]

    Jaderberg, A

    B. Jaderberg, A. A. Gentile, A. Ghosh, V. E. Elfving, C. Jones, D. Vodola, J. Manobianco, and H. Weiss, Phys- ical Review A 110, 052423 (2024)

    work page 2024

  24. [24]

    Y.-Y. Hong, C. J. E. Arce, and T.-W. Huang, IEEE Access (2023)

    work page 2023

  25. [25]

    Blazakis, Y

    K. Blazakis, Y. Katsigiannis, N. Schetakis, and G. Stavrakakis, International Conference on Frontiers of Artificial Intelligence, Ethics, and Multidisciplinary Ap- plications, , 155 (2023)

    work page 2023

  26. [26]

    Y. Y. Hong and J. B. D. Santos, IEEE Systems Journal 17, 6081 (2023)

    work page 2023

  27. [27]

    Y.-Y. Hong, C. L. P. P. Rioflorido, and W. Zhang, Ex- pert Systems with Applications 241, 122645 (2024)

    work page 2024

  28. [28]

    M. M. Sushmit and I. M. Mahbubul, Energy Conversion and Management 294, 117555 (2023)

    work page 2023

  29. [29]

    Oliveira Santos, F

    V. Oliveira Santos, F. P. Marinho, P. A. Costa Rocha, J. V. G. Th´ e, and B. Gharabaghi, Energies 17, 3580 15 (2024)

    work page 2024

  30. [30]

    Y.-Y. Hong, D. J. D. Lopez, and Y.-Y. Wang, IEEE Access (2024)

    work page 2024

  31. [31]

    Z. Kou, T. Liu, and J. Zhao, 2020 39th Chinese Control Conference (CCC), , 4318 (2020)

    work page 2020

  32. [32]

    Guilian, L

    W. Guilian, L. Tingting, T. Lu, and L. Jinlin, 2023 IEEE 2nd International Conference on Electrical Engineering, Big Data and Algorithms (EEBDA) , , 1281 (2023)

    work page 2023

  33. [33]

    Alomari and S

    A. Alomari and S. A. Kumar, 2023 IEEE Cognitive Communications for Aerospace Applications Workshop (CCAA W), , 1 (2023)

    work page 2023

  34. [34]

    R. R. Mostafa, O. Kisi, R. M. Adnan, T. Sadeghifar, and A. Kuriqi, Water 15, 486 (2023)

    work page 2023

  35. [35]

    Deepa Anbarasi and V

    J. Deepa Anbarasi and V. Radha,

    work page

  36. [36]

    Mahmud, R

    J. Mahmud, R. Mashtura, S. A. Fattah, and M. Saquib, Quantum Machine Intelligence 6, 11 (2024)

    work page 2024

  37. [37]

    Y. Gong, G. Liu, Y. Xue, R. Li, and L. Meng, Informa- tion and Software Technology 162, 107268 (2023)

    work page 2023

  38. [38]

    Budach, M

    L. Budach, M. Feuerpfeil, N. Ihde, A. Nathansen, N. Noack, H. Patzlaff, F. Naumann, and H. Harmouch, arXiv preprint arXiv:2207.14529 (2022)

    Pith/arXiv arXiv 2022

  39. [39]

    Mapa divulga os 100 munic ´ ıpios mais ricos do agroneg´ ocio em 2023,

    Minist´ erio da Agricultura, Pecu´ aria e Abastecimento (MAPA), “Mapa divulga os 100 munic ´ ıpios mais ricos do agroneg´ ocio em 2023,” (2023), accessed on: January 16, 2025

    work page 2023

  40. [40]

    Mapa municipal de barreiras - 2020,

    IBGE, “Mapa municipal de barreiras - 2020,” (2020), accessed on: December 25, 2024

    work page 2020

  41. [41]

    Prediction of Worldwide Energy Resources (POWER) Data Access Viewer,

    National Aeronautics and Space Administration (NASA), “Prediction of Worldwide Energy Resources (POWER) Data Access Viewer,” (2024), accessed on: December 25, 2024

    work page 2024

  42. [42]

    Instituto Nacional de Meteorologia (IN- MET),

    Minist´ erio da Agricultura, Pecu´ aria e Abastecimento (MAPA), “Instituto Nacional de Meteorologia (IN- MET),” (2024), accessed on: Jan. 26, 2024

    work page 2024

  43. [43]

    Portal HidroWeb - Sistema de Informa¸ c˜ oes Hidrol´ ogicas,

    Agˆ encia Nacional de ´Aguas e Saneamento B´ asico (ANA), “Portal HidroWeb - Sistema de Informa¸ c˜ oes Hidrol´ ogicas,” (2024), accessed on: December 25, 2024

    work page 2024

  44. [44]

    Ogur and I

    B. Ogur and I. Yilmaz, Quantum Inf. Comput. 23, 181 (2023)

    work page 2023

  45. [45]

    G. F. de Jesus, M. H. F. da Silva, O. M. Pires, L. C. da Silva, C. dos Santos Cruz, and V. L. da Silva, Entropy 27, 490 (2025)

    work page 2025

  46. [46]

    M. Ekman, Learning deep learning: Theory and prac- tice of neural networks, computer vision, natural lan- guage processing, and transformers using TensorFlow (Addison-Wesley Professional, 2021)

    work page 2021

  47. [47]

    Y. Li, Z. Wang, R. Han, S. Shi, J. Li, R. Shang, H. Zheng, G. Zhong, and Y. Gu, Neural Networks 166, 148 (2023)

    work page 2023

  48. [48]

    Supplementary information - weather fore- casting,

    M. H. Fraga, “Supplementary information - weather fore- casting,” (2025), gitHub repository

    work page 2025